I have a good bit of respect for the polling CNN/ORC does, having seen the behind the scenes effort that goes into it. but I see the horserace headlines in the poll and what does gets noticed is this:
1,028 adult Americans conducted by telephone by ORC International on October 14 – 17, 2015. The margin of sampling error for results based on the total sample is plus or minus 3 percentage points.
The sample also includes 956 interviews among registered voters (plus or minus 3 percentage points)
What does not get noticed in the poll is that the Republicans surveyd nationally are just 298 self-described Republicans who are registered to vote. That is not likely voters.
More so, there are an additional 167 independents who lean Republican and are registered to vote who make up the voter poll of Republicans in the poll. That portion of the poll has a margin of error of ±4.5%.
This is not an indictment of the poll, but I think it has to be an indictment of the process. We’re going to include and exclude people on a debate stage based on a poll where someone at 5% could really be at 0.5% and where someone at 0.5% could be at 5%.
That does not seem right.
It also highlights a real flaw in doing so much media coverage of these races based on horserace numbers.
And not to pick on CNN, but the NBC/Wall Street Journal poll is of “400 Republican primary voters” with a margin of error of 4.9%.
On top of that, polling of Republicans has been crap since 2010 and polling of conservatives has been crap in multiple countries since then.
I just think the press is doing it wrong.